A Novel Feature Combination Approach For Driver Fatigue Detection

Driver drowsiness is one of the major causes of traffic accidents. For instance, according to the National High Way Traffic Safety Organisation in Tunisia, 39.84% of accidents are caused by somnolence, in 2022. In that context, various state of the art methods were developed to detect distracting driving and hence prevent traffic accidents. In this paper, we propose a novel approach for driver drowsiness detection based on a new set of combined features (HOG-SURF and HOG-HARRIS). We start by an effective detection, from each frame, of the driver face. Then, we extract a Region Of Interest (ROI) containing the main expressive facial parts. Once the ROI are delimited, we extract the HOG descriptors from both HARRIS and SURF key points. The developed method was evaluated on the YAWDD database. Experimental results show the performance and competitivenes of the proposed approach. We reach an average rate of 99.15%.